In aggregate, gastrointestinal cancers account for more cancer mortality than any other organ system. While colorectal cancers are currently screened, annual US mortality from upper gastrointestinal cancers exceeds 90,000 compared to roughly 50,000 for colorectal cancer. Strikingly, 43,000 men and women are diagnosed each year with pancreatic cancer (PanC), which will cause nearly 37,000 deaths annually (Jemal et al. (2010) “Cancer statistics” CA Cancer J Clin 60: 277-300). As a result, PanC is the fourth leading cause of cancer deaths (id). Patients who present with symptoms typically already have advanced stage disease and only 15% meet criteria for potentially curative surgery (Ghaneh et al. (2007) “Biology and management of pancreatic cancer” Gut 56: 1134-52). Despite surgery, 85% will die of recurrent disease (Sohn et al. (2000) “Resected adenocarcinoma of the pancreas-616 patients: results, outcomes, and prognostic indicators” J Gastrointest Surg 4: 567-79). PanC mortality exceeds 95% and the 5-year survival rate is less than 25% for patients having curative surgery (Cleary et al (2004) “Prognostic factors in resected pancreatic adenocarcinoma: analysis of actual 5-year survivors” J Am Coll Surg 198: 722-31; Yeo et al (1995) “Pancreaticoduodenectomy for cancer of the head of the pancreas. 201 patients” Ann Surg 221: 721-33).
Among patients with syndromic predisposition to PanC or strong family history, aggressive, invasive screening strategies using computed tomography scans or endoscopic ultrasound have shown a 10% yield for neoplasia (Canto et al. (2006) “Screening for early pancreatic neoplasia in high-risk individuals: a prospective controlled study” Clin Gastroenterol Hepatol 4: 766-81). This screening strategy is impractical for the general population where most PanC arises without a known predisposition (Klein et al. (2001) “Familial pancreatic cancer” Cancer J 7: 266-73).
The nearly uniform lethality of PanC has generated intense interest in understanding pancreatic tumor biology. Precursor lesions have been identified, including pancreatic intraepithelial neoplasm (PanIN, grades I-III) and intraductal papillary mucinous neoplasm (IPMN) (Fernández-del Castillo et al. (2010) “Intraductal papillary mucinous neoplasms of the pancreas” Gastroenterology 139: 708-13, 713.e1-2; Haugk (2010) “Pancreatic intraepithelial neoplasia—can we detect early pancreatic cancer?” Histopathology 57: 503-14). Study of both precursors and malignant lesions has identified a number of molecular characteristics at genetic, epigenetic, and proteomic levels that could be exploited for therapy or used as biomarkers for early detection and screening (Kaiser (2008) “Cancer genetics. A detailed genetic portrait of the deadliest human cancers” Science 321: 1280-1; Omura et al. (2009) “Epigenetics and epigenetic alterations in pancreatic cancer” Int J Clin Exp Pathol 2: 310-26; Tonack et al. (2009) “Pancreatic cancer: proteomic approaches to a challenging disease” Pancreatology 9: 567-76). Recent tumor and metastatic lesion mapping studies have shown that there may be a significant latency period between the development of malignant PanC and the development of metastatic disease, suggesting a wide window of opportunity for detection and curative treatment of presymptomatic earliest-stage lesions (Yachida et al. (2010) “Distant metastasis occurs late during the genetic evolution of pancreatic cancer” Nature 467: 1114-7).
PanC sheds (e.g., exfoliates) cells and DNA into local effluent and ultimately into stool. For example, DNA containing a mutant KRAS gene can be identified (e.g., sequenced) from pancreatic juice of patients with pancreatic cancer, PanIN, and IPMN (Yamaguchi et al. (2005) “Pancreatic juice cytology in IPMN of the pancreas” Pancreatology 5: 416-21). Previously, highly sensitive assays have been used to detect mutant DNA in matched stools of pancreas cancer patients whose excised tumors were known to contain the same sequences (Zou et al (2009) “T2036 Pan-Detection of Gastrointestinal Neoplasms By Stool DNA Testing: Establishment of Feasibility” Gastroenterology 136: A-625). A limitation of mutation markers relates to the unwieldy process of their detection in conventional assays; typically, each mutational site across multiple genes must be assayed separately to achieve high sensitivity.
Methylated DNA has been studied as a potential class of biomarkers in the tissues of most tumor types. In many instances, DNA methyltransferases add a methyl group to DNA at cytosine-phosphate-guanine (CpG) island sites as an epigenetic control of gene expression. In a biologically attractive mechanism, acquired methylation events in promoter regions of tumor suppressor genes are thought to silence expression, thus contributing to oncogenesis. DNA methylation may be a more chemically and biologically stable diagnostic tool than RNA or protein expression (Laird (2010) “Principles and challenges of genome-wide DNA methylation analysis” Nat Rev Genet 11: 191-203). Furthermore, in other cancers like sporadic colon cancer, methylation markers offer excellent specificity and are more broadly informative and sensitive than are individual DNA mutations (Zou et al (2007) “Highly methylated genes in colorectal neoplasia: implications for screening” Cancer Epidemiol Biomarkers Prev 16: 2686-96).
Analysis of CpG islands has yielded important findings when applied to animal models and human cell lines. For example, Zhang and colleagues found that amplicons from different parts of the same CpG island may have different levels of methylation (Zhang et al. (2009) “DNA methylation analysis of chromosome 21 gene promoters at single base pair and single allele resolution” PLoS Genet 5: e1000438). Further, methylation levels were distributed bi-modally between highly methylated and unmethylated sequences, further supporting the binary switch-like pattern of DNA methyltransferase activity (Zhang et al. (2009) “DNA methylation analysis of chromosome 21 gene promoters at single base pair and single allele resolution” PLoS Genet 5: e1000438). Analysis of murine tissues in vivo and cell lines in vitro demonstrated that only about 0.3% of high CpG density promoters (HCP, defined as having >7% CpG sequence within a 300 base pair region) were methylated, whereas areas of low CpG density (LCP, defined as having <5% CpG sequence within a 300 base pair region) tended to be frequently methylated in a dynamic tissue-specific pattern (Meissner et al. (2008) “Genome-scale DNA methylation maps of pluripotent and differentiated cells” Nature 454: 766-70). HCPs include promoters for ubiquitous housekeeping genes and highly regulated developmental genes. Among the HCP sites methylated at >50% were several established markers such as Wnt 2, NDRG2, SFRP2, and BMP3 (Meissner et al. (2008) “Genome-scale DNA methylation maps of pluripotent and differentiated cells” Nature 454: 766-70).
For pancreatic cancer, PanIN, and IPMN lesions, marker methylation has been studied at the tissue level (Omura et al. (2008) “Genome-wide profiling of methylated promoters in pancreatic adenocarcinoma” Cancer Biol Ther 7: 1146-56; Sato et al. (2008) “CpG island methylation profile of pancreatic intraepithelial neoplasia” Mod Pathol 21: 238-44; Hong et al. (2008) “Multiple genes are hypermethylated in intraductal papillary mucinous neoplasms of the pancreas” Mod Pathol 21: 1499-507). For example, the markers MDFI, ZNF415, CNTNAP2, and ELOVL4 were highly methylated in 96%, 86%, 82%, and 68% of the cancers studied while, comparatively, only 9%, 6%, 3%, and 7% of control (non-cancerous) pancreata, respectively, were highly methylated at these same four loci (Omura et al. (2008) “Genome-wide profiling of methylated promoters in pancreatic adenocarcinoma” Cancer Biol Ther 7: 1146-56). It was found that measuring the methylation state of both MDFI and CNTNAP2 provided an indicator for pancreatic cancer that had both a high sensitivity (>90%) and a high specificity (>85%) (Omura et al. (2008) “Genome-wide profiling of methylated promoters in pancreatic adenocarcinoma” Cancer Biol Ther 7: 1146-56).
Furthermore, Sato and colleagues found eight genes differentially expressed in pancreatic cancer cell lines before and after treatment with a methyltransferase inhibitor (Sato et al. (2003) “Discovery of novel targets for aberrant methylation in pancreatic carcinoma using high-throughput microarrays” Cancer Res 63: 3735-42). These markers were subsequently assessed by methylation-specific PCR (MSP) of DNA from Pan-IN lesions. The results showed that SARP-2 (secreted frizzled related protein 1, SFRP1) had 83% sensitivity and could discriminate between Pan-IN 2 and higher grade Pan-IN 3 (Sato et al. (2008) “CpG island methylation profile of pancreatic intraepithelial neoplasia” Mod Pathol 21: 238-44). Discrimination of a high grade precursor or early stage cancer from a lower grade lesion is important when considering the morbidity of pancreaticoduodenectomy or distal pancreatectomy, the main surgical therapies for PanC. When studying both main-duct and side-branch IPMN precursors, Hong and colleagues showed high sensitivity and specificity for SFRP1 as well, especially in combination with UCHL1 (Hong et al. (2008) “Multiple genes are hypermethylated in intraductal papillary mucinous neoplasms of the pancreas” Mod Pathol 21: 1499-507). Tissue factor pathway inhibitor 2 (TFPI2) has a well-established tumor suppressor role in GU and GI malignancies, including prostate, cervical, colorectal, gastric, esophageal, and pancreatic cancers (Ma et al. (2011) “MicroRNA-616 induces androgen-independent growth of prostate cancer cells by suppressing expression of tissue factor pathway inhibitor TFPI-2” Cancer Res 71: 583-92; Lim et al. (2010) “Cervical dysplasia: assessing methylation status (Methylight) of CCNA1, DAPK1, HS3ST2, PAX1 and TFPI2 to improve diagnostic accuracy” Gynecol Oncol 119: 225-31; Hibi et al. (2010) “Methylation of TFPI2 gene is frequently detected in advanced well-differentiated colorectal cancer” Anticancer Res 30: 1205-7; Glockner et al. (2009) “Methylation of TFPI2 in stool DNA: a potential novel biomarker for the detection of colorectal cancer” Cancer Res 69: 4691-9; Hibi et al. (2010) “Methylation of the TFPI2 gene is frequently detected in advanced gastric carcinoma” Anticancer Res 30: 4131-3; Tsunoda et al. (2009) “Methylation of CLDN6, FBN2, RBP1, RBP4, TFPI2, and TMEFF2 in esophageal squamous cell carcinoma” Oncol Rep 21: 1067-73; Tang et al. (2010) “Prognostic significance of tissue factor pathway inhibitor-2 in pancreatic carcinoma and its effect on tumor invasion and metastasis” Med Oncol 27: 867-75; Brune et al. (2008) “Genetic and epigenetic alterations of familial pancreatic cancers” Cancer Epidemiol Biomarkers Prev 17: 3536-4). This marker has also been shown to be shed into the GI lumen and was 73% sensitive when assayed from pancreatic juice of cancers and normal subjects (Matsubayashi et al. (2006) “DNA methylation alterations in the pancreatic juice of patients with suspected pancreatic disease” Cancer Res 66: 1208-17).
TFPI2 was among a large number of potential mutation and methylation markers studied in tissue and stool samples as candidates for colorectal neoplasia. In a training-test set analysis of archival stools from almost 700 subjects, a multi-marker methylation panel, including TFPI2, BMP3, NDRG4, and vimentin was shown to have 85% sensitivity in CRC and 64% sensitivity in advanced colorectal adenomas, both at 90% specificity (Ahlquist D et al. (2010) “Next Generation Stool DNA Testing for Detection of Colorectal Neoplasia—Early Marker Evaluation”, presented at Colorectal Cancer: Biology to Therapy, American Association for Cancer Research).
Previous research has tested the performance of colorectal cancer methylation markers in PanC detection. In particular, a case-control study compared DNA from PanC tumor cases to DNA from colonic epithelia using MSP targeting markers previously reported in PanC (e.g., MDFI, SFRP2, UCHL1, CNTNAP2, and TFPI2) and additional discriminant colonic neoplasm markers (e.g., BMP3, EYA4, Vimentin, and NDRG4). A multi-marker regression model showed that EYA4, UCHL1, and MDFI were highly discriminant, with an area under the receiver operating characteristics curve of 0.85. As an individual marker, BMP3 was found to have an area under the receiver operator characteristics curve of 0.90. These four markers and mutant KRAS were subsequently assayed in a larger set of stool samples from PanC subjects in a blinded comparison to matched stools from individuals with a normal colonoscopy. Individually, BMP3 and KRAS were highly specific but poorly sensitive; in combination, sensitivity improved to 65% while maintaining 88% specificity (Kisiel, et al. (2011) “Stool DNA screening for colorectal cancer: opportunities to improve value with next generation tests” J Clin Gastroenterol 45: 301-8. These results suggested that methylation differences in UCHL1, EYA4, and MDFI at the level of the pancreas were obscured by background colonic methylation in the stool-based comparison. As such, cancer screening is in need of a marker or marker panel for PanC that is broadly informative and exhibits high specificity for PanC at the tissue level when interrogated in samples taken from a subject (e.g., a stool sample).